Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
UNCLASSIFIED
UNCLASSIFIED: Distribution Statement A. Approved for public release; distribution is unlimited
TARDEC Ground Vehicle Robotics:
Vehicle Dynamic Characterization and Research
by
Joseph Selikoff
Report Documentation Page Form ApprovedOMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering andmaintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, ArlingtonVA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if itdoes not display a currently valid OMB control number.
1. REPORT DATE 01 SEP 2015 2. REPORT TYPE
3. DATES COVERED 00-00-2015 to 00-00-2015
4. TITLE AND SUBTITLE TARDEC Ground Vehicle Robotics: Vehicle Dynamic Characterizationand Research
5a. CONTRACT NUMBER
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) Joseph Selikoff
5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) US Army RDECOM-TARDEC,6501 E. 11 Mile Road,Warren,MI,48397-5000
8. PERFORMING ORGANIZATIONREPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT See Report
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF ABSTRACT Same as
Report (SAR)
18. NUMBEROF PAGES
40
19a. NAME OFRESPONSIBLE PERSON
a. REPORT unclassified
b. ABSTRACT unclassified
c. THIS PAGE unclassified
Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18
UNCLASSIFIED
UNCLASSIFIED
Table of Contents
I. Introduction
II. Physical Vehicle Design Properties
a. All-Terrain Vehicle Classification
b. UGV Classification
c. OEMs in Military Vehicle Engineering
d. General Vehicle Designs and Specifications
III. Definitions and Numerical Simulations of Vehicle Characteristics
a. Vehicle Mobility
b. Analytical Resources
c. Demonstration of Vehicle Parameter-Dynamic Handling Relationships
IV. Experimental Methods for Vehicle Characterization
a. Experimental Methods to Estimate Vehicle Mobility
b. Hardware to Estimate Parameters On-line
c. Examples of Experimental Data Analysis
UNCLASSIFIED
UNCLASSIFIED
Introduction
Vehicle dynamics, especially for military applications, has become a highly researched
and published topic. Starting from the creation of the first automobiles, the analysis techniques
for the dynamics of these vehicles has been developed constantly. Once computers became
commonplace, the modelling capabilities of researchers increased dramatically. The goal of this
paper is to introduce and define some military and civilian vehicle types that are common, detail
some of the methods used to numerically evaluate vehicle mobility, and provide an overview for
some of the more common methods and hardware used to experimentally determine vehicle
characteristics.
Physical Vehicle Design Properties
The physical components of a vehicle naturally have a large effect on its performance.
Body and chassis type have the largest effect on overall vehicle mass, while things like
suspension design and drive train construction can greatly affect the lateral and longitudinal
capabilities of a vehicle. This section will list vehicle classifications, vehicle design choices and
how they affect characteristics important to vehicle dynamics, and original equipment
manufacturers (OEMs) and contractors that are involved in the military vehicle market.
All-Terrain Vehicle Classification
The following is a list and description of the different classifications of military armed
vehicles described in the Treaty on Conventional Armed Forces in Europe [12]:
UNCLASSIFIED
UNCLASSIFIED
Tanks: Tracked and armored vehicles weighing at least 16.5 metric tons unladen*, with a main
gun larger than 75mm caliber that can rotate 360 degrees about the body of the vehicle.
Wheeled vehicles that meet the above criteria will also be deemed as tanks.
Armored Combat Vehicles: Self-propelled vehicle with armor protection and cross-country
capability. These vehicles include the following:
Armored Personnel Carrier: An armored combat vehicle designed to transport combat
infantry and is armed with either a mounted (crew serviced) weapon or hand held
weapon of less than 20mm caliber.
Armored Infantry Fighting Vehicle: An armored combat vehicle designed and
equipped to transport a combat infantry squad, and which provides the capability
for the troops to deliver fire from inside the vehicle protected by armor. This
vehicle must be armed with either a mounted (crew serviced) weapon or hand
held weapon of at least 20mm caliber and sometimes an antitank missile
launcher. This vehicle serves as the principle weapon system for armored,
mechanized, and motorized infantry formations.
Heavy Armament Combat Vehicle: An armored combat vehicle with a gun of at least
75mm caliber, weighing at least 6 metric tons unladen*, and which does not fall
into any of the classifications for armored personnel carrier, armored infantry
fighting vehicle, or battle tank.
*Unladen Weight: Weight of vehicle excluding ammunition, fuel, oil, removable armor, spare
parts, tools, accessories, snorkeling equipment, crew, and personal equipment.
UNCLASSIFIED
UNCLASSIFIED
UGV Classification
There are two main situations that unmanned vehicles are considered a better
alternative to manned operations. First, and considered most important by many, is any
dangerous situation that could otherwise injure or kill someone. A great example of this is a
bomb squad robot. More often than not, these robots are used to assess a bomb site to decide
how to deal with the charge. Instead of having a soldier or civilian officer inspect the bomb by
hand, driving the robot up remotely keeps these people much further from the blast zone. The
second situation where unmanned ground vehicles are being developed is in jobs that can
induce human error based on fatigue or stress. An example of this is autonomous long haul
convoy movement, where vehicles travel in packs for hundreds of miles at a time by following a
leader vehicle with technologies like machine vision and GPS waypoint setting and following.
The following is a list and description of the three main types of unmanned ground vehicles:
Remote Controlled Operation: An unmanned vehicle that is operated by a human. This term
specifically describes a system which must remain within the line-of-sight of the operator,
because there is no camera feedback. This can be limiting for stealth and long distance
operations, but can be a cheaper solution if the situation does not require stealth, as in
sacrificial mine clearing or something of that nature.
Teleoperation: An unmanned vehicle that is operated by a human. This term specifically
describes a system which does not necessarily need to remain within the line-of-sight of
the operator to function. To achieve this, a camera is mounted on the vehicle and the
feed is transmitted back to the operator to be monitored. This can be used for many
applications, namely long distance or building searching operations. This is the
preferred method of human operated robotics because the operators are able to be
much farther away from the situation the robot is being sent in to.
UNCLASSIFIED
UNCLASSIFIED
Autonomous Operation: An unmanned vehicle that operates based on its own sensors and
control laws without the real time input of a human operator. These vehicle systems are
becoming more prevalent in recent years. The system uses its sensors to develop a
certain awareness of the environment around it. With this awareness, control algorithms
are used to have the vehicle make decisions about what actions to take. These vehicles
can be used as scouts, follow waypoints to make resupply an unmanned operation, and
develop new paths through terrain that are more efficient and safer for manned convoys
to travel.
OEMs in Military Vehicle Engineering
Many government contracts involving ground vehicles are outsourced, in part or in full, to
private manufacturers and contractors. This helps the military get the most out of their
investment because they can choose who gets signed to work on a project based on bidding
price, along with previous work done by the companies. The following is a list of a few well
known OEMs and contractors, along with some of the military vehicle projects they have worked
on, most of which can be found within 10 to 15 miles from the TARDEC facility in Detroit:
Oshkosh Defense:
Light Tactical: JLTV, HMMWV, L-ATV, S-ATV
Medium Tactical: Family of Medium Tactical Vehicles (FMTV)
Heavy Tactical: Heavy Expanded Mobility Tactical Truck (HEMTT), Heavy Equipment
Transporter (HET)
Mine-Resistant Ambush Protected (MRAP): M-ATV
UNCLASSIFIED
UNCLASSIFIED
Vehicle Systems: CORE 1080 (Crew protection, perception, and survivability system),
TAK-4 (Independent Suspension System, applicable to many light and medium weight
tactical vehicles, including the complete MRAP series), PROPULSE (Hybrid Diesel-
Electric System with Export Power), Command Zone (integrated vehicle control and
diagnostic system), and TerraMax (Unmanned ground vehicle technology)
Lockheed Martin Corporation
They work on many different types of vehicles, but have only one ground vehicle project
in the works. The JLTV is the one ground vehicle project that Lockheed Martin is
working on, and they are working in a partnership with BAE Systems.
BAE Systems
Amphibious Combat Vehicles
Armored Multi-Purpose Vehicle (AMPV): Tank without a turret
CV90: Tank
BR90: Mobile bridging system.
Bradley Fighting Vehicle: Tank
RG33 Mine-Resistant, Ambush Protected Vehicle (MRAP)
General Dynamics Corporation
Abrams Tank: Main Battle Tank
Light Armored Vehicles (LAV): Stryker and LAV family of vehicles
Mine-Resistant, Ambush Protected Vehicle (MRAP) Family
UNCLASSIFIED
UNCLASSIFIED
AM General, LLC
Blast Resistant Vehicle-Off Road (BRV-O): Their JLTV offering
High Mobility Multi-purpose Wheeled Vehicle (HMMWV/Humvee): Classical canvas
version and up-armored version.
Modernized Light Tactical Vehicle (MLTV)
M-1165 2.0 Deployable Reconnaissance Ground Network-Vehicle (DRGN-V)
General Vehicle Designs and Specifications
This section lists different types of components and system types for a few different
subassemblies that would be common on ground vehicles.
Powertrain Systems: Gas Powered, Diesel, Turbo Diesel, Gas Turbine, Hybrid: Gas-
Electric, Diesel-Electric, Series, Parallel.
Power Distribution: RWD, FWD, AWD, open diff, LSD, Torsen diff, differential braking
(traction control), drive by wire electric, Electric traction control.
Suspension Styles: Suspension is what keeps the vehicle off the ground and
mechanically isolated from the terrain irregularities. The suspension usually consists of
a spring component and a damping component attached to points on both the main
chassis of the vehicle and a mounting point on the suspension linkage. Different types
of suspension have niches in different situations. Sometimes a dependent suspension is
much more suitable, while other times an independent suspension is much more
suitable for the vehicle’s needs.
UNCLASSIFIED
UNCLASSIFIED
Dependent: I-Beam, Panhard Rod, Leaf Spring, satchel link, watt’s linkage,
WOB Link, Mumford Linkage.
Independent: Swing Axle, Sliding Pillar, MacPherson Strut, Double Wishbone,
Multi-link, Semi trailing arm, Swing Arm, Leaf Springs
Tire Choices: Most vehicles are going to have some type of all-terrain tires that will be
able to traverse most anything effectively. The size and strength of these tires depends
more on the weight and performance capabilities of the vehicle they are on. Some type
of run-flat tires should be mentioned, but they are usually much heavier than classic
pneumatic tires. Also, bead-locking rims could be quite beneficial for certain
applications, especially those where lower tire pressure would be necessary to get
through a certain path, like in extremely rocky or muddy terrain.
Definitions and Numerical Simulations of Vehicle Characteristics
Being able to objectively assess the mobility of a vehicle is critical in its design
and development process. Without being able to characterize the performance of a
vehicle, critical metrics that define a vehicle’s motion would be hard to know accurately,
and therefore meaningful design changes would be hard to verify without physically
constructing prototypes. In this section, general definitions of mobility, along with
mathematical expressions that provide a basis for vehicle dynamic simulation, will be
discussed. Also, demonstrations of how vehicle parameters affect the performance of a
vehicle are presented. The relationship between the vehicle and the terrain it is
traversing is mentioned, but not deeply discussed.
UNCLASSIFIED
UNCLASSIFIED
Vehicle Mobility
Vehicle mobility in military applications is beginning to be better defined by standards like
the NATO Reference Mobility Model (NRMM), which is the standard in the Army Battle
Command, Simulation, and Experimentation Directorate for single vehicle ground movement
[13]. These standards continue to be developed and defined by the Army Corps of Engineers,
whose development of computer models for vehicle mobility is generalized to accommodate
multiple vehicle platforms, unlike previous models that needed to be redesigned for different
parameters [13].
Essentially, mobility is the ability of a vehicle to maneuver and navigate over a specified
terrain as it is influenced and changed by weather and other environmental conditions [3].
There are multiple ways to define the ever- changing vehicle-terrain dynamic model, such as the
Nepean Wheeled/Tracked Vehicle Performance Model. Demand for engineers and designers in
both the private and the military sector is becoming more and more evident, as companies and
government agencies realize the importance of computer simulation in the prototyping process
[14]. Studies detailing the effects of torque distribution on the tractive ability of vehicles have
also been conducted in the hopes of finding a way to actively tune traction control software to
the current terrain [11].
Along with the analysis of single vehicles, there is growing interest in the study of the
collective dynamics of a group of vehicles. This analysis can involve the effect of new tread
marks in terrain on a follower vehicle’s dynamics, and can also detail how communication
between the vehicles within a convoy can allow follower vehicles to alter power distribution to
make the most out of available traction [6], [10].
The easier part of the analysis of a vehicle’s mobility is finding the dynamic
characteristics of a vehicle on an assumed flat surface. The hard part is accounting for terrain,
UNCLASSIFIED
UNCLASSIFIED
especially natural and/or dynamic terrain that can change as it is being driven on. There should
be a direct correlation between terrain “severity” and the ability of a vehicle to maneuver through
it. Some models solve this issue by concocting a terrain factor that will essentially be used to
modify the ideal model of the vehicle [7]. This can only serve as an estimation, and because of
this, sensor data and dynamic modelling with sensor input is invaluable to aid in the
understanding of how the characteristic handling of a vehicle change on different terrain.
Terrain characteristics can be defined with testing like cone penetrometers, measuring
the force needed to push a metal cone into the ground a certain distance [14]. Different types of
soil can also be defined by factors like soil density and yield of oat as it is compacted by multiple
vehicle passes [6]
Analytical Resources
Traction and tire slippage are two main criteria for mobility, because all of the ground
forces are transferred to the vehicle from the ground via the tires. Therefore, traction is usually
the factor, more than any other factors, which limits the vehicle’s mobility. Some other criteria
are longitudinal top speed, acceleration, and braking capabilities.
Another component of vehicle mobility that is very important is the ability to create yaw
moment and therefore yaw acceleration, which translates to a lateral acceleration of the vehicle.
This factor can be characterized by something called an understeer gradient, which is the
difference of the theoretical ratios of the front and rear weights divided by the tire slip angle
being experienced in either the front or the back. It can also be described as the ratio of steer
angle over lateral acceleration. This gradient essentially describes how a vehicle would react as
it travelled around a certain radius turn with a continuous acceleration. A positive understeer
gradient means that the vehicle is understeer, a negative understeer gradient means that the
UNCLASSIFIED
UNCLASSIFIED
vehicle is oversteer, and an understeer gradient of zero means the vehicle is neutral steer.
Understeer means that as the vehicle velocity is increased, the steer angle necessary to stay on
the radius of the turn will increase. Oversteer means that as the vehicle velocity is increased,
the steer angle necessary to stay on the radius of the turn will decrease. Neutral steer means
that as the vehicle velocity is increased, the steer angle necessary to stay on the radius of the
turn will be constant. [5]
An important factor in low speed vehicle cornering has to do with turning radius and
steer angles. There is a relation called Ackermann steering, which is simply a relation between
the inside and outside tire angles in a turn. This also has to do with wheel base length. The
modelling of the physical body movement of the vehicle is not too difficult. The hard part is
defining and characterizing the forces the tires can give out. This characterization can be done
with something called the Pacejka method, or The Magic Formula. This formula is literally
derived from Hans looking at experimental data and trying to make a mathematical curve fit the
experimental data as closely as possible. Many vehicle dynamics simulations use this curve to
define the tire forces. Sometimes though, in simpler models, the tire curve can be defined as a
line that ends at the peak force of the tire. This is obviously inaccurate, but is sufficient for
certain calculations. [5] Some other tire models will be described later.
Equations: (** - From Gillespie [5])
Low Speed Maneuvering **: Unlike steady state maneuvers at speed, tires don’t need to
generate lateral forces in low speed or “parking lot” maneuvers. Because of this, they roll with
no slip angles. This means that the steering angles of the front wheels must be calibrated
perfectly to allow a no slip condition while turning. The steer angles can be expressed as
𝛿𝛿𝑜𝑜 =𝐿𝐿
𝑅𝑅 +𝑡𝑡𝑓𝑓2
UNCLASSIFIED
UNCLASSIFIED
and
𝛿𝛿𝑖𝑖 =𝐿𝐿
𝑅𝑅 −𝑡𝑡𝑓𝑓2
With the average front wheel angle, known as the Ackerman Angle, is given by
𝛿𝛿 =𝐿𝐿𝑅𝑅
assuming small angles.
Tire Models [2]: Tire models take in slip angles and output a lateral force generated by that slip
angle. The simplest model applies a strictly linear relationship between slip angle and lateral
force. There are two other models that are quite common: The Pacejka Model, and the Duggoff
Model.
Pacejka Model: The Pacejka model was developed essentially by taking experimental data and
developing an equation that fit the curve of the data as closely as possible. The problem with
this is that the coefficients of the equations change for each tire type, and the only way to know
is to either run experiments to find the coefficients or look up tables and try to find the tire you
are using on a surface similar to what you want to analyze. There are many, many different
things that can affect the specific values for these curves, but they are nonetheless good
estimates. The Pacejka model is defined as follows:
𝐷𝐷 = 𝑎𝑎1𝐹𝐹𝑧𝑧3 + 𝑎𝑎2𝐹𝐹𝑧𝑧2 + 𝑎𝑎3𝐹𝐹𝑧𝑧
𝐵𝐵𝐵𝐵𝐷𝐷 = 𝑎𝑎4sin (𝑎𝑎5 tan−1(𝑎𝑎6𝐹𝐹𝑧𝑧))
𝐸𝐸 = 𝑎𝑎7𝐹𝐹𝑧𝑧2 + 𝑎𝑎8𝐹𝐹𝑧𝑧 + 𝑎𝑎9
𝐵𝐵 =𝐵𝐵𝐵𝐵𝐷𝐷𝐵𝐵𝐷𝐷
UNCLASSIFIED
UNCLASSIFIED
𝜑𝜑𝑃𝑃𝑃𝑃𝑃𝑃 = (1 − 𝐸𝐸)𝛼𝛼 +𝐸𝐸𝐵𝐵
tan−1(𝐵𝐵𝛼𝛼)
𝐹𝐹𝑦𝑦 = 𝐷𝐷 sin (𝐵𝐵 tan−1(𝐵𝐵𝜑𝜑𝑃𝑃𝑃𝑃𝑃𝑃))
where a1 – a9 and C are variables found from experimental data, A, B, D,E, and 𝜑𝜑𝑃𝑃𝑃𝑃𝑃𝑃 are
calculated values with no physical meaning, Fz is the normal force on the tire, and 𝛼𝛼 is
the tire slip angle.
The shape of a standard Pacejka curve at different normal loads can be seen below.
Duggoff Model: The Duggoff tire model is probably the most practical
tire model available. It is so useful because it can be defined based only on physical
characteristics, instead of the crazy polynomial fit coefficients that the Pacejka model
uses. It is also more accurate than a linear model because it flattens out after a certain
slip angle, similar to the Pacejka curve. The Duggoff Model can be expressed as
𝜆𝜆 =𝜇𝜇𝑝𝑝𝑝𝑝𝐹𝐹𝑧𝑧
2𝐵𝐵𝛼𝛼|tan(𝛼𝛼)|
0 10 20 30 40 50 60 70 80 900
1000
2000
3000
4000
5000
6000
7000Tire Curves for Various Normal Loads
Slip Angle Alpha (deg.)
Late
ral F
orce
(N)
2 kN4 kN6 kN8 kN
UNCLASSIFIED
UNCLASSIFIED
𝑓𝑓(𝜆𝜆) = �(2 − 𝜆𝜆)𝜆𝜆 𝑖𝑖𝑓𝑓 𝜆𝜆 < 11 𝑖𝑖𝑓𝑓 𝜆𝜆 > 1
𝐹𝐹𝑦𝑦 = 𝐵𝐵𝛼𝛼 tan(𝛼𝛼)𝑓𝑓(𝜆𝜆)
where 𝜇𝜇𝑝𝑝𝑝𝑝 is the peak coefficient of friction between the tire and the ground.
The Duggoff curve can be seen below at various different normal loads.
Below is a comparison between the Duggoff and Pacejka curves. They are relatively
close.
UNCLASSIFIED
UNCLASSIFIED
Steady-State Cornering **: This segment is performed under the bicycle model assumptions
that the difference between the inside and outside slip angles in the front and rear, along with
the steer angles, is small enough to be neglected. This means there is one set of angles per
axle. Things like camber thrust, and the kinematic effects of the suspension on camber and
steer angles, are neglected in these equations.
Lateral Axle Force **:
𝐹𝐹𝑦𝑦𝑦𝑦 = 𝑀𝑀𝑏𝑏𝐿𝐿 �
𝑉𝑉2
𝑅𝑅 �
𝐹𝐹𝑦𝑦𝑓𝑓 = 𝐹𝐹𝑦𝑦𝑦𝑦𝑐𝑐𝑏𝑏
where Fyf is the lateral force on the front axle, Fyr is the lateral force on the rear axle, M is the
vehicle mass, b is the distance from the cg to the front axle, c is the distance from the cg to the
rear axle, L is the total wheelbase length, V is the longitudinal velocity, and R is the radius of the
turn.
UNCLASSIFIED
UNCLASSIFIED
Steer Angle **:
𝛼𝛼𝑓𝑓 = 𝑊𝑊𝑓𝑓𝑉𝑉2
𝐵𝐵𝛼𝛼𝑓𝑓𝑔𝑔𝑅𝑅
𝛼𝛼𝑦𝑦 = 𝑊𝑊𝑦𝑦𝑉𝑉2
𝐵𝐵𝛼𝛼𝑦𝑦𝑔𝑔𝑅𝑅
𝛿𝛿 = 57.3𝐿𝐿𝑅𝑅
+ 𝛼𝛼𝑓𝑓 − 𝛼𝛼𝑦𝑦 ⇒ 57.3𝐿𝐿𝑅𝑅
+ �𝑊𝑊𝑓𝑓
𝐵𝐵𝛼𝛼𝑓𝑓−𝑊𝑊𝑦𝑦
𝐵𝐵𝛼𝛼𝑦𝑦�𝑉𝑉2
𝑔𝑔𝑅𝑅
𝐾𝐾 = 𝑊𝑊𝑓𝑓
𝐶𝐶𝛼𝛼𝑓𝑓− 𝑊𝑊𝑟𝑟
𝐶𝐶𝛼𝛼𝑟𝑟
𝑎𝑎𝑦𝑦 = 𝑉𝑉2
𝑔𝑔𝑔𝑔
𝛿𝛿 = 57.3 𝐿𝐿𝑔𝑔
+ 𝐾𝐾𝑎𝑎𝑦𝑦
where 𝛼𝛼𝑓𝑓 and 𝛼𝛼𝑦𝑦 are the front and rear slip angles, Wf and Wr are the weights on
the front and rear axle, 𝐵𝐵𝛼𝛼𝑓𝑓 and 𝐵𝐵𝛼𝛼𝑦𝑦 are the front and rear tire cornering stiffness
per axle, 𝛿𝛿 is the steer angle in degrees, K is the understeer gradient, and ay is
the lateral acceleration in g’s.
Characteristic and Critical Speeds **:
Characteristic speed is defined as the speed at which the required steer angle to
hold the turn radius is twice the Ackerman angle for an understeer vehicle. It is given as
𝑉𝑉𝑃𝑃ℎ𝑃𝑃𝑦𝑦 = �57.3 𝐿𝐿𝑔𝑔𝐾𝐾
Critical speed is defined as the speed at which an oversteer vehicle will become
unstable. It is given as
UNCLASSIFIED
UNCLASSIFIED
𝑉𝑉𝑃𝑃𝑦𝑦𝑖𝑖𝑐𝑐 = �−57.3 𝐿𝐿𝑔𝑔𝐾𝐾
These two definitions imply an important difference between understeer and oversteer
vehicles that can be defined when the transfer functions of the systems are analyzed. An
understeer vehicle can never go unstable, aka eigenvalues in the right half-plane. But, at a certain
point the steering inputs have no effect on the dynamics of the vehicle. On the other hand, an
oversteer vehicle can definitely become unstable, but the steering inputs always have effect on
the vehicle dynamics. This highlights the reason why most consumer automobiles come from the
factory in an understeer configuration, so that if something goes wrong in a maneuver, the car will
not go crazy and spin out, and the likely overreaction of the driver wouldn’t make the situation
necessarily worse. Also, it makes sense now why race teams usually set up their vehicle with
neutral steer leaning towards oversteer, so that the driver can maintain control as he or she
operates the vehicle near its dynamics limits.
Roll Angle and Weight Transfer **:
𝐻𝐻 = ℎ𝑃𝑃𝑔𝑔 − �ℎ𝑓𝑓 + ℎ𝑦𝑦
2 �
𝑟𝑟𝑟𝑟 = 𝑘𝑘𝜑𝜑𝑓𝑓 + 𝑘𝑘𝜑𝜑𝑦𝑦 −𝑊𝑊𝐻𝐻
𝜑𝜑 =𝑊𝑊𝐻𝐻𝑟𝑟𝑟𝑟
∗ 𝑎𝑎𝑦𝑦
∆𝐹𝐹𝑧𝑧,𝑓𝑓 =𝑘𝑘𝜑𝜑𝑓𝑓𝜑𝜑 + 𝑊𝑊𝑓𝑓ℎ𝑓𝑓𝑎𝑎𝑦𝑦
𝑡𝑡𝑓𝑓
UNCLASSIFIED
UNCLASSIFIED
∆𝐹𝐹𝑧𝑧,𝑦𝑦 =𝑘𝑘𝜑𝜑𝑦𝑦𝜑𝜑 + 𝑊𝑊𝑦𝑦ℎ𝑦𝑦𝑎𝑎𝑦𝑦
𝑡𝑡𝑦𝑦
𝐹𝐹𝑓𝑓,𝑖𝑖 =𝑊𝑊𝑓𝑓
2− ∆𝐹𝐹𝑧𝑧,𝑓𝑓
𝐹𝐹𝑓𝑓,𝑜𝑜 =𝑊𝑊𝑓𝑓
2+ ∆𝐹𝐹𝑧𝑧,𝑓𝑓
𝐹𝐹𝑦𝑦,𝑖𝑖 =𝑊𝑊𝑦𝑦
2− ∆𝐹𝐹𝑧𝑧,𝑦𝑦
𝐹𝐹𝑦𝑦,𝑖𝑖 =𝑊𝑊𝑦𝑦
2+ ∆𝐹𝐹𝑧𝑧,𝑦𝑦
where H is the effective roll moment arm from the ground, hcg is the height of the center of
gravity from the ground, hf and hr are the height of the front and rear suspension roll centers
from the ground, 𝑘𝑘𝜑𝜑𝑓𝑓 and 𝑘𝑘𝜑𝜑𝑦𝑦 are the front and rear roll stiffness, rr is the roll rate, 𝜑𝜑 is the roll
angle, tf and tr are the front and rear track widths, ∆𝐹𝐹𝑧𝑧,𝑓𝑓 and ∆𝐹𝐹𝑧𝑧,𝑦𝑦 are the weight transfer in the
front and rear, and 𝐹𝐹𝑓𝑓,𝑖𝑖, 𝐹𝐹𝑓𝑓,𝑜𝑜, 𝐹𝐹𝑦𝑦,𝑖𝑖, and 𝐹𝐹𝑦𝑦,𝑖𝑖 are the front and rear inner and outer vertical loads on
the tires.
Transient Maneuvering Dynamics [15]: This segment is going to skip all of the derivation of
this model because it is pretty intense. I will do my best to mention every assumption, but there
are a lot to make this happen. The system can be modelled as follows:
𝐵𝐵0 = 𝐵𝐵𝛼𝛼𝑓𝑓 + 𝐵𝐵𝛼𝛼𝑦𝑦
𝐵𝐵1 = 𝑎𝑎𝐵𝐵𝛼𝛼𝑓𝑓 − 𝑏𝑏𝐵𝐵𝛼𝛼𝑦𝑦
𝐵𝐵2 = 𝑎𝑎2𝐵𝐵𝛼𝛼𝑓𝑓 + 𝑏𝑏2𝐵𝐵𝛼𝛼𝑦𝑦
UNCLASSIFIED
UNCLASSIFIED
�
�̇�𝑟𝑉𝑉�̇�𝑦𝑟𝑟𝑉𝑉𝑦𝑦
� =
⎣⎢⎢⎢⎢⎡
−𝐵𝐵2𝑉𝑉𝑉𝑉
−𝐵𝐵1𝑉𝑉𝑉𝑉
0 0
−𝐵𝐵1 −𝑀𝑀𝑉𝑉2
𝑀𝑀𝑉𝑉−𝐵𝐵0𝑀𝑀𝑉𝑉
0 01 0 0 00 1 𝑉𝑉 0⎦
⎥⎥⎥⎥⎤
�
𝑟𝑟𝑉𝑉𝑦𝑦𝜃𝜃𝑦𝑦
�+
⎣⎢⎢⎢⎢⎡𝑎𝑎𝐵𝐵𝛼𝛼𝑓𝑓𝑉𝑉𝐵𝐵𝛼𝛼𝑓𝑓𝑀𝑀00 ⎦
⎥⎥⎥⎥⎤
𝛿𝛿(𝑐𝑐)
𝑌𝑌 = [1 0 0 0] �
𝑟𝑟𝑉𝑉𝑦𝑦𝜃𝜃𝑦𝑦
�
There are a few new variables in this that we haven’t seen yet. V is the longitudinal velocity, but
is being assumed as essentially the velocity magnitude to simplify the state equations. r is the
yaw rate. Vy is the lateral velocity, and J is the yaw moment of inertia of the vehicle. This model
assumes small angles, uses the bicycle model, and uses a modified linear tire model. Side slip
angle, or the angle between the velocity vector at the cg of the vehicle and the longitudinal axis
of the vehicle, can be found with
𝛽𝛽 = tan−1 �𝑉𝑉𝑦𝑦𝑉𝑉�
and is very important in the development of traction control algorithms.
Extra Tidbits **: There are a few different little equations that either fit in all of the above
situations are standalone equations that shed light on different factors of vehicle dynamics.
Wheel Slip: If there is an acceleration, there will always be wheel slip. Some example data is
shown in part 3. The equation to compute wheel slip is
%𝑆𝑆𝑆𝑆𝑖𝑖𝑆𝑆 =𝑉𝑉 − 𝑅𝑅𝑤𝑤𝜔𝜔
𝑉𝑉∗ 100%
where 𝜔𝜔 is the rotational speed of the wheel and Rw is the wheel radius.
UNCLASSIFIED
UNCLASSIFIED
Yaw Velocity Gain: This equation is pretty cool because you can use it to find KUS while the
vehicle is doing any type of maneuver, instead of just doing the skid pad test, if a few vehicle
states are known. This can be expressed as
𝑟𝑟𝛿𝛿
=𝑉𝑉/𝐿𝐿
1 + 𝐾𝐾𝑈𝑈𝑈𝑈𝑉𝑉257.3 𝐿𝐿𝑔𝑔
this is a little complicated, but if yaw rate, steer angle, and velocity are being measured and the
wheelbase length is known, the understeer gradient can be found without too much effort.
Understeer Increment: This is essentially steering system slop. This has a bit to do with
suspension design, which isn’t something I want to dive into, so I’ll keep the explanations brief
and google can fill in the blanks. The expression for understeer increment is given as
𝐾𝐾𝑈𝑈𝑈𝑈,𝑠𝑠𝑐𝑐𝑦𝑦𝑔𝑔 =𝑊𝑊𝑓𝑓(𝑅𝑅𝑤𝑤𝜈𝜈 + 𝑆𝑆)
𝐾𝐾𝑈𝑈𝑈𝑈
where 𝐾𝐾𝑈𝑈𝑈𝑈,𝑠𝑠𝑐𝑐𝑦𝑦𝑔𝑔 is the understeer increment due to steering, 𝜈𝜈 is the Caster angle, or the angle
that the wheel rotates around as it is steered when looked at from the side, p is the pneumatic
trail, or distance the tire patch is trailing the center of the wheel, and KSS is the spring coefficient
between the steering wheel and the wheel on the road.
UNCLASSIFIED
UNCLASSIFIED
Demonstration of Vehicle Parameter-Dynamic Handling Relationships
The following are a few examples of results from mathematical simulations. There are
parameters varied between simulations to demonstrate how the changes affect vehicle
performance in simulation:
Zero to Sixty Simulation
UNCLASSIFIED
UNCLASSIFIED
Standard Parameters
Twice the Mass
UNCLASSIFIED
UNCLASSIFIED
Half the Mass
100 meter Skid Pad Simulation
Standard Parameters
UNCLASSIFIED
UNCLASSIFIED
Double the Mass
Half the Mass
UNCLASSIFIED
UNCLASSIFIED
Drastically Rear Heavy
Drastically Front Heavy
UNCLASSIFIED
UNCLASSIFIED
Forward Angled Slope between Roll Center Heights
Rearward Facing Slope between Roll Center Heights
Lateral Acceleration (g's)0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Stee
r Ang
le (d
eg.)
1
2
3
4Understeer Gradient
Time(Sec)0 5 10 15 20 25
Stee
r Ang
le (d
eg.)
1
2
3
4Steer Angle over Time
UNCLASSIFIED
UNCLASSIFIED
Double Lane Change Maneuver
Standard Parameters
Double the Weight
UNCLASSIFIED
UNCLASSIFIED
Half the Weight
Experimental Methods for Vehicle Characterization
While simulated vehicle testing is extremely useful to aid in the design process, the
product eventually has to be made, and the designs need to be validated. That is where
experimental real time data logging comes in. While the sensing technologies used today still
aren’t perfect, they help shine a light on the dynamic vehicle characteristics. In this section,
testing techniques, sensors utilized in these tests, and examples of some of the pitfalls of on-
line data logging, are described.
Experimental Methods to Estimate Vehicle Mobility
Skid Pad: Starting from rest, a vehicle accelerates slowly around a constant radius turn
until the car loses traction and skids out. This test is essential in characterizing the
UNCLASSIFIED
UNCLASSIFIED
steady state cornering capabilities of the vehicle. When actually done to a real vehicle, it
is important that this test is done going both clockwise and counterclockwise around the
course.
Double Lane Change: Unlike the skid pad test, most vehicle handling and maneuvers
are not steady state; most involve transient weight transfer and conditions. The double
lane change is a standard road vehicle test, and can be used both to asses a car’s
handling abilities and the abilities of an autonomous control system. The test involves
the vehicle starting going straight forward, swerving over the distance that two lanes
would entail, and swerving back the other direction the same distance. The maneuver is
transient because the sudden change in direction induces body roll, and therefore weight
transfer which affects the traction of the vehicle [4].
UNCLASSIFIED
UNCLASSIFIED
Acceleration and Deceleration: This is pretty straight forward, a car starts at rest and
accelerates in a straight line for a certain distance, then slams on the brakes until the car
comes back to rest. This is a test of a vehicles longitudinal acceleration capabilities.
Coast Down: This test helps to characterize the air drag, rolling resistance, and/or
inertial and viscous losses in the drivetrain, depending on how the vehicle is set up.
Hardware to Estimate Parameters On-Line
Wheel Speed Sensors
Usually Hall Effect Sensors, much less often rotary digital encoders. Encoders
are much better at low speed readings, Hall Effect sensors are more reliable for high
speed readings. Hall Effect sensors are usually bad for low speed because most of
these types of sensors use spokes of some kind, on a structure called a reluctor, to
create a signal. If the sensor is between spokes, there is no reading to be made. So
between spoke sensor impulses, the speed change is unknown.
UNCLASSIFIED
UNCLASSIFIED
IMU and Processors
Accelerometers and gyroscopic sensors are used to measure the transient
motion of the vehicle. Gyros are used primarily to find yaw rate. In fact, some purpose
made vehicle dynamic IMUs, or inertial measurement units, have only a z-axis gyro, as
yaw rate is so important for vehicle dynamic characterization. Accelerometers are useful
in all three axis, though they can be misleading if certain precautions aren’t taken.
Usually the accelerometers are used in a vehicle to find vehicle pitch and roll angles.
This helps to characterize vehicle weight transfer. This can be tricky though, because as
the vehicle rolls, the lateral acceleration seen in the y-axis gyro is sensing both lateral
acceleration and some gravitational acceleration because the vehicle is rolling. It is
extremely important that the code used to process the dynamic information from these
sensors accounts for body roll in the calculation of lateral acceleration and body pitch in
the calculation of longitudinal acceleration.
GPS
GPS sensors aren’t accurate enough to give premium vehicle dynamic data yet,
but it can be used as a check against the integral of accelerometers to check that the
system is calibrated correctly.
Suspension Position Sensors
One way to physically verify the position of the suspension to analyze the
kinematic properties of the system through different maneuvers. This can be used to
both validate the kinematic design software used to design the system, and also can
help verify the inferred roll angles that are found with the IMU. This is usually done with
UNCLASSIFIED
UNCLASSIFIED
linear potentiometers, which have an electrical resistance that changes at a set rate as
the length of the device is changed.
Steering Position Sensors
Helps validate the mathematical relationship between steering input, velocity, and
vehicle parameters. These are usually rotary digital encoders, with one part mounted on
the column support, and the other end rotating with the steering column. This
measurement can be extrapolated through the steering rack ratio to get the input
steering angle at the wheels, which is an important measurement in vehicle dynamic
analysis.
Examples of Experimental Data Analysis
A comparison between velocity and engine RPM measurements from a vehicle’s CAN
bus as it does a quarter mile race, and the simulation of the same event, can be seen below.
There are many different reasons the simulation and experimental data may not line up.
Initialization of the simulation may be inaccurate. There may be inherent bias in the sensors.
Even though the experimental data has been filtered, there still is likely some small trace of the
noise that was present in the raw data.
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
This is another example of some data gathered from a vehicle’s CAN bus. This is just
the wheel speeds of all four wheels. It is interesting to look at how the wheel speeds vary when
the vehicle is taking a turn, as seen in the first plot below. In the second picture, data was
collected while slamming on the brakes in loose gravel. It is clear that the ABS is failing for the
front tires, but the rear tires are well controlled and do not skid.
UNCLASSIFIED
UNCLASSIFIED
Below are plots that are essentially the starting point for an estimation system. Data was
collected for a vehicle doing some random maneuvers, and the steer angle and velocity data
was then input into a simulation model to compare some states like slip angle and yaw rate and
verify the simulation model. In the plots below it is evident that the simulation tracks with the
experimental data quite well.
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
References
1. Brudnak, Mark, Patrick Nunez, and Alexander Reid. Real-time, Distributed, Unmanned Ground Vehicle Dynamics and Mobility Simulation. Technical rept. no. 2002-01-1178. N.p.: SAE, 2002. Print. SAE Technical Paper Series.
2. Daily, Robert, and William Travis. Cascaded observers to improve lateral vehicle state and
tyre parameter estimates. Ed. David M. Bevly. N.p.: n.p., n.d. Print.
3. Dattathreya, Macam, and Harpreet Singh. "A Novel Approach for Combat Vehicle Mobility Definition and Assessment." SAE World Congress (2012): n. pag. Print.
4. Forkenbrock, Garrick J., comp. Public Workshop on Dynamic Rollover and Handling Test
Techniques. 3 Dec. 2002, Vehicle Research and Test Center. N.p.: n.p., n.d. Print.
5. Gillespie, Thomas D. Fundamentals of Vehicle Dynamics. Warrendale: SAE, 1992. Print.
6. Gray, Jeremy Patrick. "Unmanned Ground Vehicle Terrain/Road Mobility Foundation and Its Algorithmic Formalization for Wheel Power Management to Improve Leader and Follower Capabilities." Diss. The University of Alabama at Birmingham, 2015. Print.
7. Hegazy, Shawky, and Corina Sandu. "Experimental investigation of vehicle mobility using a novel wheel mobility number." Journal of Terramechanics 50 (2013): 303-10. Print.
8. Hutangkabodee, Suksun, et al. "Soil Parameter Identification for Wheel-terrain Interaction
Dynamics and Traversability Prediction." International Journal of Automation and Computing 3 (2006): 244-51. Print.
9. Joo, Sang Hyun, et al. "Real time traversability analysis to enhance rough terrain navigationo
for a 6 x 6 autonomous vehicle." Journal of Mechanical Science and Technology 27.4 (2013): 1125-34. Print.
10. Senatore, C., and C. Sandu. "Off-road tire modeling and the multi-pass effect for vehicle
dynamics simulation." Journal of Terramechanics 48 (2011): 265-76. Print.
11. - -. "Torque distribution influence on tractive efficiency and mobility of off-road wheeled vehicles." Journal of Terramechanics 48 (2011): 372-83. Print.
12. Treaty on Conventional Armed Forces in Europe, Vienna, 19 November 1990, NATO
Summit, available from http://www.osce.org/library/14087?download=true
13. US Army Corps of Engineers: ERDC/GSL. Standard for Ground Vehicle Mobility. By E. Alex Baylot, Jr. et al. Technical rept. no. TR-05-6. N.p.: n.p., 2005. Print.
UNCLASSIFIED
UNCLASSIFIED
14. Wong, J.Y. Computer-Aided Methods for Performance and Design Evaluation of Off-Road Vehicles from the Traction Perspective. Technical rept. no. 2011-01-0266. N.p.: SAE, 2011. Print.
15. Zhang, Yuan. Drive Point Mobility, Transmissibility and Beyond. Technical rept. no. 2011-
01-0502. N.p.: SAE, 2011. Print.